parser.py

# parser.py
import ast

def get_category(node):
    """Determine the category of an AST node."""
    if isinstance(node, (ast.Import, ast.ImportFrom)):
        return 'import'
    elif isinstance(node, (ast.Assign, ast.AnnAssign, ast.AugAssign)):
        return 'assignment'
    elif isinstance(node, ast.FunctionDef):
        return 'function'
    elif isinstance(node, ast.AsyncFunctionDef):
        return 'async_function'
    elif isinstance(node, ast.ClassDef):
        return 'class'
    elif isinstance(node, ast.If):
        return 'if'
    elif isinstance(node, ast.While):
        return 'while'
    elif isinstance(node, ast.For):
        return 'for'
    elif isinstance(node, ast.Expr):
        return 'expression'
    else:
        return 'other'

def create_vector(category, level, location, total_lines):
    """Create a vector representation for a code part."""
    # Vector: [category_id, level, start_line_normalized, end_line_normalized]
    category_map = {
        'import': 1, 'assignment': 2, 'function': 3, 'async_function': 4, 'class': 5,
        'if': 6, 'while': 7, 'for': 8, 'expression': 9, 'spacer': 10, 'other': 11
    }
    category_id = category_map.get(category, 0)
    start_line, end_line = location
    return [
        category_id,
        level,
        start_line / total_lines,  # Normalized start position
        end_line / total_lines     # Normalized end position
    ]

def parse_node(node, lines, prev_end, level=0, total_lines=None):
    """Recursively parse an AST node and its children, assigning hierarchy levels."""
    if total_lines is None:
        total_lines = len(lines)

    parts = []
    start_line = getattr(node, 'lineno', prev_end + 1)
    end_line = getattr(node, 'end_lineno', start_line)

    # Handle spacers before the node
    if start_line > prev_end + 1:
        spacer_lines = lines[prev_end:start_line - 1]
        spacer_vector = create_vector('spacer', level, (prev_end + 1, start_line - 1), total_lines)
        parts.append({
            'category': 'spacer',
            'source': ''.join(spacer_lines),
            'location': (prev_end + 1, start_line - 1),
            'level': level,
            'vector': spacer_vector
        })

    # Capture the node's source
    stmt_lines = lines[start_line - 1:end_line]
    node_vector = create_vector(get_category(node), level, (start_line, end_line), total_lines)
    parts.append({
        'category': get_category(node),
        'source': ''.join(stmt_lines),
        'location': (start_line, end_line),
        'level': level,
        'vector': node_vector
    })

    # Process nested nodes (e.g., class/function bodies, control structures)
    if hasattr(node, 'body'):
        nested_prev_end = end_line - 1
        for child in node.body:
            child_parts = parse_node(child, lines, nested_prev_end, level + 1, total_lines)
            parts.extend(child_parts)
            nested_prev_end = child_parts[-1]['location'][1]

    # Handle additional bodies (e.g., else, elif, orelse for loops)
    if hasattr(node, 'orelse') and node.orelse:
        orelse_prev_end = parts[-1]['location'][1]
        for child in node.orelse:
            child_parts = parse_node(child, lines, orelse_prev_end, level + 1, total_lines)
            parts.extend(child_parts)
            orelse_prev_end = child_parts[-1]['location'][1]

    return parts

def parse_python_code(code):
    """Parse Python code string and return parts with hierarchy and vectors."""
    lines = code.splitlines(keepends=True)
    total_lines = len(lines)
    try:
        tree = ast.parse(code)
    except SyntaxError:
        return [{'category': 'error', 'source': 'Invalid Python code', 'location': (1, 1), 'level': 0, 'vector': [0, 0, 1.0, 1.0]}]

    parts = []
    prev_end = 0

    for stmt in tree.body:
        stmt_parts = parse_node(stmt, lines, prev_end, total_lines=total_lines)
        parts.extend(stmt_parts)
        prev_end = stmt_parts[-1]['location'][1]

    # Capture trailing spacers
    if prev_end < total_lines:
        remaining_lines = lines[prev_end:]
        spacer_vector = create_vector('spacer', 0, (prev_end + 1, total_lines + 1), total_lines)
        parts.append({
            'category': 'spacer',
            'source': ''.join(remaining_lines),
            'location': (prev_end + 1, total_lines + 1),
            'level': 0,
            'vector': spacer_vector
        })

    return parts